Production of extremely low volatile organiccompounds from biogenic emissions: Measuredyields and atmospheric implicationsTuija Jokinena,b,1, Torsten Berndta, Risto Makkonenb, Veli-Matti Kerminenb, Heikki Junninenb, Pauli Paasonenb,Frank Stratmanna, Hartmut Herrmanna, Alex B. Guentherc, Douglas R. Worsnopb,d, Markku Kulmalab, Mikael Ehnb,and Mikko Sipiläb

aLeibniz-Institut für Troposphärenforschung (TROPOS), 04318 Leipzig, Germany; bDepartment of Physics, University of Helsinki, 00014 Helsinki, Finland;cAtmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, Richland, WA 99354; and dAerodyne Research, Inc., Billerica, MA 01821

Edited by John H. Seinfeld, California Institute of Technology, Pasadena, CA, and approved April 28, 2015 (received for review December 16, 2014)

Oxidation products of monoterpenes and isoprene have a majorinfluence on the global secondary organic aerosol (SOA) burdenand the production of atmospheric nanoparticles and cloud con-densation nuclei (CCN). Here, we investigate the formation ofextremely low volatility organic compounds (ELVOC) from O3 andOH radical oxidation of several monoterpenes and isoprene in aseries of laboratory experiments. We show that ELVOC from allprecursors are formed within the first minute after the initial attackof an oxidant. We demonstrate that under atmospherically relevantconcentrations, species with an endocyclic double bond efficientlyproduce ELVOC from ozonolysis, whereas the yields from OH radi-cal-initiated reactions are smaller. If the double bond is exocyclic orthe compound itself is acyclic, ozonolysis produces less ELVOC andthe role of the OH radical-initiated ELVOC formation is increased.Isoprene oxidation produces marginal quantities of ELVOC regard-less of the oxidant. Implementing our laboratory findings into aglobal modeling framework shows that biogenic SOA formationin general, and ELVOC in particular, play crucial roles in atmosphericCCN production. Monoterpene oxidation products enhance atmo-spheric new particle formation and growth in most continental re-gions, thereby increasing CCN concentrations, especially at highvalues of cloud supersaturation. Isoprene-derived SOA tends to sup-press atmospheric new particle formation, yet it assists the growthof sub-CCN-size primary particles to CCN. Taking into account com-pound specific monoterpene emissions has a moderate effect on themodeled global CCN budget.

autoxidation | ELVOC | monoterpenes | isoprene | new particle formation

Formation and subsequent growth of new aerosol particles is amajor source of cloud condensation nuclei (CCN) in the

global troposphere (1, 2), and a big contributor to the largereported uncertainty in the radiative forcing by aerosol−cloudinteractions (3–7). Multiple field studies have shown that CCNproduction is tightly connected with the oxidation of biogenicvolatile organic compounds (BVOC) emitted by terrestrial eco-systems (8–11). To explain these observations, large-scale modelsimulations demonstrate a need for a BVOC oxidation mecha-nism in the atmosphere that produces very low volatility organicvapors with molar formation yields of at least a few percent perreacted precursor compound (12–14).The existence and formation mechanisms of essentially non-

volatile organic vapors in the atmosphere have puzzled scientistsfor some time (14–16). Such extremely low volatile organic com-pounds (ELVOC) (17) were recently detected, both in laboratorystudies and in the ambient atmosphere (18), yet typical atmo-spheric oxidation chemistry schemes do not explain ELVOCproduced on a time scale of minutes or hours. Furthermore,current state-of-the-art models using available chemistry schemeshave systematically failed to reproduce the observed concentra-tions and volatility of organic aerosol components (17, 19). A

plausible explanation for the fast ELVOC production was recentlygiven by Ehn et al. (20), who proposed that highly oxidized (O:C ≈1) ELVOC are formed as first-generation oxidation products ofα-pinene, a monoterpene emitted in vast quantities by differentecosystems. The authors proposed that α-pinene oxidation byozone (O3) forms peroxy radicals (RO2), which undergo successiveintramolecular hydrogen shifts followed by a rapid reaction withO2, resulting in prompt production of high levels of oxygenation.This formation pathway, leading to highly oxygenated RO2 radicaland ELVOC formation, was more recently confirmed by Jokinenet al. (21) and Rissanen et al. (22). They showed that autoxidation,a mechanism known to be important in the liquid phase and whichhas been hypothesized to take place also in the gas phase (23),could mechanistically explain the gas-phase ELVOC formation.Their results also unambiguously demonstrated that highly oxi-dized RO2 radicals and closed-shell ELVOC monomers (with aC10 carbon skeleton) and dimers (C20 carbon skeleton) are formedon time scales of seconds, thereby indicating immediate pro-duction of condensable vapors close to emission sources.Surprisingly, according to Ehn et al. (20), α-pinene appears

to produce ELVOC with a much higher molar yield fromozonolysis (∼7%) than from the OH radical reaction (<1%).Jokinen et al. (21) investigated the relative importance of O3and OH radicals in the formation of highly oxygenated RO2from limonene and α-pinene, both having a reactive endocyclic

Significance

Extremely low volatility organic compounds (ELVOC) are sug-gested to promote aerosol particle formation and cloud conden-sation nuclei (CCN) production in the atmosphere. We show thatthe capability of biogenic VOC (BVOC) to produce ELVOC dependsstrongly on their chemical structure and relative oxidant levels.BVOC with an endocyclic double bond, representative emissionsfrom, e.g., boreal forests, efficiently produce ELVOC from ozo-nolysis. Compounds with exocyclic double bonds or acyclic com-pounds including isoprene, emission representative of the tropics,produce minor quantities of ELVOC, and the role of OH radicaloxidation is relatively larger. Implementing these findings into aglobal modeling framework shows that detailed assessment ofELVOC production pathways is crucial for understanding biogenicsecondary organic aerosol and atmospheric CCN formation.

Author contributions: T.J., T.B., R.M., V.-M.K., F.S., H.H., D.R.W., M.K., M.E., and M.S.designed research; T.J., T.B., R.M., and A.B.G. performed research; H.J. and A.B.G. con-tributed new reagents/analytic tools; T.J., T.B., R.M., H.J., P.P., A.B.G., M.E., and M.S.analyzed data; and T.J., T.B., R.M., V.-M.K., and M.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed: Email: tuija.jokinen@helsinki.fi.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1423977112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1423977112 PNAS | June 9, 2015 | vol. 112 | no. 23 | 7123–7128

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Dec

embe

r 28

, 202

0

double bond. They found that the amount of highly oxidizedRO2 radicals formed from the OH radical oxidation was only asmall fraction of that from the pure ozonolysis reaction. Theobserved ELVOC yield from β-pinene ozonolysis in Ehn et al.(20) was much lower than that from α-pinene ozonolysis, in-dicating that the results obtained for α-pinene and limonenecannot be generalized to all BVOC, or even to all monoterpenes,present in the atmosphere. The relative yields of ELVOC andmore volatile oxidized organics are key parameters for both re-gional and global CCN budgets, as their influence on atmo-spheric new particle formation as well as growth of both newlyformed particles and sub-CCN-size primary particles can bemarkedly different (e.g., refs. 3 and 12).The mixture of BVOC emitted to the atmosphere contains a

large number of compounds with different chemical structures(24). The magnitude and variability of these emissions are, how-ever, not well understood. Even in the boreal forest, composed ofonly a few dominant tree species, BVOC emission patterns arecomplex, with prominent seasonal cycles (25) and large differenceseven between individual trees of the same species (26). In thetropics, with more diverse vegetation properties, relatively fewBVOC emission assessments have been published so far (e.g., refs.27 and 28). Future BVOC emission characteristics are expected tochange considerably in many regions as a result of increasingtemperatures and CO2 concentrations, as well as changes in ex-treme weather conditions and vegetation cover (25, 29–31).In this work, we investigate ELVOC formation and associated

CCN production from BVOC species having different chemicalstructures. We determine ELVOC yields from both ozonolysisand OH radical-initiated oxidation of five major BVOC speciesthrough comprehensive laboratory experiments. These speciesinclude monoterpenes with endocyclic and exocyclic doublebonds and acyclic compounds with reactive double bonds (here-after referred as endocyclic, exocyclic, and acyclic terpenes) as wellas isoprene. Based on the observed ELVOC yields, we then useglobal model simulations to investigate the implications of ourfindings on atmospheric CCN production.

Results and DiscussionExperimental Findings. We conducted extensive laboratory ex-periments of ELVOC formation in an atmospheric pressure flowtube (TROPOS-LFT) using five differently structured BVOC:two endocyclic (limonene and α-pinene) and one exocyclic (β-pinene)and one acyclic (myrcene) monoterpene as well as isoprene, whichtogether represent up to 70% of the global biogenic hydrocarbonemissions (24). We detected rapid formation of highly oxygenatedspecies arising from repetitive addition of O2 in the course ofoxidation after initial reaction of the terpene with O3 and OHradicals with all studied terpenes (see SI Appendix for experi-mental conditions). Since the vapor pressures of the formed oxi-dation products cannot be measured experimentally, the vaporpressures were estimated with three different methods (17, 32, 33).Monoterpene oxidation products with formula of C10H16O≥7 werethus classified as ELVOC (see SI Appendix for details).In the case of endocyclic monoterpenes, closed-shell ELVOC

compounds (C10H14O7, C10H14O9, and C10H14O11) dominatedthe spectra together with the RO2 radical ELVOC species(C10H15O8 and C10H15O10) (Fig. 1A and SI Appendix, Fig. S3A).In limonene experiments, ozonolysis products with C9 carbonskeleton were also detected, but with lower intensities. This isbecause the oxidation reaction also takes place in the second,less reactive, double bond. Acyclic myrcene, containing threedouble bonds, created the most-versatile ELVOC products, withprominent features of C10 and some C7 monomers (Fig. 1B andSI Appendix, Fig. S1C). The experiment with β-pinene showedfeatures from OH reactions in the form of C10 monomers andweaker ozonolysis signals, the C9 monomers (SI Appendix, Fig.S3B). Altogether, the ELVOC signals from β-pinene were ∼50

times lower than those from α-pinene oxidation. Dimers (formedin RO2 + RO2 reactions) were detected from all monoterpenes,and they corresponded mostly to C20, and to some extent other,such as C19, compounds. In the case of myrcene, some lower-intensity dimer signals corresponding to C10, C14, and C17 com-pounds were also observed (Fig. 1B).Isoprene produced very low amounts of ELVOC (SI Appendix,

Figs. S1 and S3C). To unambiguously identify the isoprene oxi-dation products from other possible background peaks, we alsostudied the oxidation of 13C labeled isoprene (2-methyl-1,3-butadiene-1-13C). We identified several highly oxygenated C5species (RO2 radicals and closed-shell products) that were clas-sified as ELVOC because of their low vapor pressures due to theshort carbon chain with very high oxygen content, which conse-quently has implications for their role in aerosol formation. Inthis study, only the isoprene monomer species with formula ofC5H8O≥8 were classified as extremely low volatility species, makingthe total molar yield of ELVOC remarkably low. Isoprene dimerformation (C10 compounds) was only detected with extremely highconcentrations of isoprene, and monoterpene contamination hin-dered the determination of dimers. Thus, the isoprene ELVOCyield represents a low-end estimate since dimer formation had to beneglected completely. However, finding an ELVOC-producingpathway is a vital discovery, since isoprene is the most emittednonmethane BVOC globally.In Fig. 1, we show the time series of selected RO2 radicals and

closed-shell monomer ELVOC from oxidation experiment oflimonene and myrcene. Limonene yields relatively high con-centrations of both RO2 radicals and closed-shell ELVOC, whilemyrcene, an acyclic compound, produces ELVOC much lessefficiently (see SI Appendix, Table S2, for experimental condi-tions). After the signals reached steady-state behavior, we con-ducted experiments where synthetic air, normally used as carriergas, was substituted by a N2-rich carrier gas. The substantialdecrease in the O2 concentration (1/60 O2 content of ambientair) reduced ELVOC production dramatically for all of thestudied terpenes (Fig. 1 and SI Appendix, Fig. S3). These ob-servations support the previous findings of Jokinen et al. (21),and we conclude that all of the studied terpenes are able to formhighly oxygenated, low-volatility reaction products via an au-toxidation mechanism under atmospheric conditions.Fig. 1 also shows how ELVOC signals are affected by reduced

OH radical concentrations (OH radicals are unavoidably produced

Fig. 1. Effect of low oxygen level and OH scavenger (propane, C3H8, or hy-drogen, H2) on ELVOC production during oxidation of (A) limonene and (B)myrcene. During low-oxygen experiments, ELVOC formation is significantlysuppressed as O2 addition to the intermediate reaction products becomesslower. During limonene oxidation, only a few ELVOC species are affected byOH scavenging, illustrating that ozonolysis is the main pathway for ELVOCformation in our experiment. In the case of myrcene, a clear decrease in totalELVOC was observed upon OH scavenging, demonstrating the role of OHoxidation in the ELVOC formation pathway. C10H16O7 and C10H14O7 areidentified as dimers from small ozonolysis-produced RO2 radicals.

7124 | www.pnas.org/cgi/doi/10.1073/pnas.1423977112 Jokinen et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 28

, 202

0

in alkene ozonolysis) in the flow tube in the case of endocycliclimonene and exocyclic myrcene. Substantially decreasing the OHradical levels allowed us to determine whether the origin of eachELVOC peak in the mass spectrum was from ozonolysis or OH-initiated reaction. From these studies (see also SI Appendix, Fig.S3), we conclude that ozonolysis has a dominant, and OH-initiatedoxidation a minor, role in the ELVOC production from endocyclicterpenes in our experiments. For the acyclic or exocyclic terpenes—myrcene, β-pinene, and isoprene—the results differ substantially,with relatively higher ELVOC yields from the OH radical-initiatedoxidation compared with ozonolysis. Our results indicate that ozo-nolysis of endocyclic alkenes is very efficient in producing ELVOC,confirming and further extending the findings by Ehn et al. (20),Rissanen et al. (22), and Jokinen et al. (21).Terpene oxidation over a wide range of experimental reactant

concentrations (atmospheric and higher) showed a nearly linearrelationship between the total ELVOC concentration and theamount of reacted terpenes (Fig. 2). The slight curvatures of theslopes of Fig. 2 likely reflect an increase in the ELVOC dimerformation with an increase in reactant consumption due to theincreasing dimerization of ELVOC RO2 radicals. The totalELVOC concentrations given in Fig. 2 depend on both O3 andOH radical reactions. The oxidant specific molar yields, Y(O3)for pure ozonolysis and Y(OH) for OH radical-initiated oxida-tion, summarized in Table 1 were calculated based on experi-ments with and without an OH scavenger (Fig. 1 and SIAppendix, Fig. S3; see SI Appendix for details on yield calcula-tions and uncertainty estimates). Our yield data represent lower-limit estimates for monoterpenes, yet our yield of 3.4 ± 1.7%Y(O3) from the ozonolysis of α-pinene is in reasonable agree-ment with the previous result of 7 ± 4% by Ehn et al. (20).Similar results have also been reported for a commonly usedsurrogate for α-pinene, cyclohexene, with an ozonolysis ELVOCyield of 4.5 ± 3.8% (22).

Global Model Simulations. To find out how ELVOC influenceatmospheric CCN production, we conducted a series of globalmodel simulations (see Table 2, Materials and Methods, and SIAppendix). We fixed the total secondary organic aerosol (SOA)yield from BVOC oxidation but varied the set of compoundscapable of producing SOA and ELVOC. We assumed that

ELVOC influence both new particle formation and growth, andthat all of them condense onto preexisting particles according tothe Fuchs-corrected aerosol surface area, as one would expectfor any extremely low volatile compound (12, 17). We tookmonoterpene ELVOC yields directly from the measurements(Table 1) and assumed zero yield for isoprene. The rest of theSOA, produced from semivolatile organic compounds (SVOC),was partitioned according to preexisting organic mass over theaerosol size distribution, reflecting thermodynamic gas−particlepartitioning typical for semivolatile organic aerosol (12). Weconcentrated on three quantities: the total particle numberconcentration (CN) and CCN concentrations at the supersatu-rations of 0.2%, CCN(0.2%), and 1%, CCN(1.0%). For typicalatmospheric aerosol populations, CCN(1.0%) and CCN(0.2%)correspond roughly to the total number concentration of parti-cles of >50 nm and >100−200 nm in diameter, respectively (2).As a result, CN and CCN(1.0)% are expected to be sensitive toatmospheric new particle formation and growth, and therebysensitive to ELVOC, whereas CCN(0.2%) is influenced greatlyby the rest of SOA that effectively partitions into preexistingprimary particles.The total annual emissions of isoprene, endocyclic, and other

monoterpenes were 519 Tg·y−1, 45 Tg·y−1, and 31 Tg·y−1, re-spectively, in our simulations. The highest contribution fromendocyclic to total monoterpenes was found in boreal andequatorial forests (SI Appendix, Fig. S4). While isoprene emis-sions exceeded total monoterpene emissions by a factor of ∼5,monoterpenes dominated total BVOC emissions at high lati-tudes in the northern hemisphere throughout most of the year.By assuming no SOA formation from NOx oxidation and totalSOA yields of 15% and 5% for monoterpenes and isoprene,respectively, we obtained an annual SOA formation of 7 Tg frommonoterpenes and 20 Tg from isoprene. While the total SOAformation rate of 27 Tg·y−1 is higher than that in most AeroCommodels (34), it is still considerably lower than both AMS-con-strained [50−380 Tg·y−1 (35)] and top-down [88 TgC·y−1 (19)]estimates. In our control experiment (CTRL) that included bothmonoterpene and isoprene oxidation (see Table 2), mono-terpene oxidation was responsible for 26% the total annual SOAformation, and 3% of the SOA originated from ELVOC.The control simulation showed a strong land−ocean contrast

in CCN concentrations, ranging from 100 cm−3 over remoteoceans to 2,000−3,000 cm−3 in continental areas (Fig. 3, firstrow). Polar areas had very low CCN concentrations of a few tensper cubic centimeter, and even continental areas north of 60°Nhad typical CCN concentrations below a few hundred per cubiccentimeter. The simulated spatial CCN distribution is in agree-ment with the recent multimodel assessment by Mann et al. (36).We found large spatial differences in how SOA formation affectsthe submicron particle population and thus CCN concentrations(Fig. 3, second row). Most notably, SOA formation increasedCCN(0.2%) over almost all continental areas and decreasedboth CN and CCN(1.0%) throughout most of North Americaand Europe. The main reason for this is that even though ad-ditional SOA tends to enhance the CCN activity of small primaryparticles, it suppresses new particle formation and growth by

Fig. 2. Total ELVOC concentrations as a function of reacted terpene (O3 andOH radical reaction). The experimental molar yields are calculated from theslopes. Uncertainty of all of the measurements is −50/+100%. Reactionconditions are given in SI Appendix.

Table 1. Total molar yields of ELVOC from selected terpeneoxidation with uncertainty of −50/+100% arising fromcalibration

Yield Limonene α-Pinene Myrcene β-Pinene Isoprene

Y (O3), % 5.3 (1.56) 3.4 (1) 0.47 (0.14) 0.12 (0.035) 0.01 (0.003)Y (OH), % 0.93 (0.27) 0.44 (0.13) 1.0 (0.29) 0.58 (0.17) 0.03 (0.009)

The data in brackets represent ELVOC yields relative to the ozonolysis ofα-pinene.

Jokinen et al. PNAS | June 9, 2015 | vol. 112 | no. 23 | 7125

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Dec

embe

r 28

, 202

0

increasing the rate at which condensable vapors and newlyformed particles are lost to the preexisting particle population.According to our simulations, ELVOC enhance new particle

formation and growth everywhere (Fig. 3, third row), therebyincreasing total CN concentrations throughout most of thecontinental areas. This is reflected also in CCN(1.0%), whichshowed clear increases over boreal forests, South America, andIndonesia. The influence of ELVOC on CCN(0.2%) was rela-tively minor, both in the absolute and relative senses. The effectof monoterpene speciation on aerosol population remained mod-erate in our simulations.Since isoprene does not produce ELVOC in our simulations, its

oxidation is likely to suppress new particle formation and growth.To confirm this effect, an additional simulation was madewhere isoprene-derived SOA was completely omitted. As shownby Fig. 3 (fourth row), isoprene SOA decreased CN andCCN(1.0%) concentrations over the continents between a fewpercent and about 20%. However, isoprene SOA enhanced thegrowth of primary particles, which can be seen as the increaseof CCN(0.2%) in many regions.Regionally, the effects of SOA and, especially, ELVOC on

CCN are highly sensitive to the preexisting aerosol size distribu-tion, the ratio between monoterpene and isoprene emissions, andthe fraction of endocyclic monoterpenes. SI Appendix, Table S4,

summarizes the simulated budgets of SOA, CN, and CCN as wellas new particle growth rates and condensation sinks globally, andseparately over four subregions located in Siberia, Amazon,Australia, and the United States during the summer season (seeFig. 3, first panel). We can see that SOA formation increases CNand CCN concentrations in Siberia and Amazon, mainly be-cause of the very clear enhancing effect of ELVOC on newparticle formation and growth in these relatively unpollutedregions. In the United States and Australia, with more intensiveprimary particle emissions, SOA formation causes an increasein CCN(0.2%) but a decrease in both CN and CCN(1.0%). Thecontribution of ELVOC to the SOA formation varied from 1.6%to 5.6% between the four subregions. ELVOC increased CN andCCN in all these regions, and, in terms of CCN(1.0%), this in-crease was 5%, 6%, 2%, and 1% in Siberia, Amazon, Australia,and the United States, respectively. The rather small fractionalincrease of CCN(1.0%) in the United States is mainly because ofthe high CCN concentration levels in that area, as the absoluteincrease in CCN(1.0%) due to ELVOCs was very similar be-tween the United States and Siberia. The importance of BVOCspeciation (MTAPINENE, in which all monoterpenes weretreated as α-pinene, see Table 2) on new particle growth and therebyon CN is clear in Siberia and Amazon, but the resulting changesin CCN concentrations remained well below 5%. Most of the

Table 2. Description of simulation experiments

Experiment Monoterpenes Isoprene ELVOC SVOC Total SOA production, Tg Remarks

CTRL yes yes yes yes 27 controlNOSOA no no no no 0 no SOA formationNOISOP yes no yes yes 7 no isopreneMTAPINENE yes yes yes yes 27 all monoterpenes as α-pineneNOELVOC yes yes no yes 27 ELVOC assumed as SVOC

The columns indicate which precursors and SOA production pathways are included in the experiment.

Fig. 3. Annual average CN, CCN(1.0%), and CCN(0.2%) concentrations in control experiment (CTRL, top row), and the relative effects of SOA, ELVOC, andisoprene. Note the different color scales in each panel. Four regions for detailed analysis are indicated in the first panel.

7126 | www.pnas.org/cgi/doi/10.1073/pnas.1423977112 Jokinen et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 28

, 202

0

SOA in the four subregions originates from isoprene. Sinceisoprene does not produce ELVOC in our model simulations,the dominant effect of isoprene SOA is to increase the organicmass of preexisting primary particles and to suppress new par-ticle formation and growth via the increased condensation sink.This latter effect is consistent between all of the four regions:When isoprene was omitted from the model, the condensationsink decreased by 4−8%, leading to higher values of CN andCCN(1.0%). However, CCN(0.2%) appears to be lower in theabsence of isoprene SOA due to the less efficient growth of smallprimary particles.To our knowledge, the model developed in this study is the

first global aerosol model that combines a hybrid SOA formationmechanism (kinetic and mass-based) with ELVOC yields obtainedfrom laboratory studies. However, the model excludes severalimportant SOA formation pathways and related processes. Themodel assumes fixed total SOA yields from monoterpene andisoprene oxidation and instantaneous nonreversible condensa-tion of BVOC oxidation products. These features affect thevertical and horizontal distribution of SOA as well as the timescales of SOA formation, and do not allow us to investigate howchanges in ambient temperature or relative humidity, atmo-spheric oxidation capacity, or preexisting particle populationwould influence the total amount of SOA. The anthropogenicinfluence on simulated SOA formation is limited to prescribedNOx fields and the aerosol distribution from anthropogenicsources. In reality, the anthropogenic modification could explaina major fraction of total SOA formation. In addition, the currentstudy omits the anthropogenic SOA precursors, which couldexplain ∼10 Tg of the annual SOA formation. The model includesno heterogenous SOA formation, which would add to the totalamount of SOA and influence its spatial and temporal distribution.The overall effect of the excluded processes is likely toward in-creased sink and suppression of new particle formation and growth.

Summary and Atmospheric Implications.We studied the productionof ELVOC from O3 and OH radical oxidation of endocyclic,exocyclic, and acyclic monoterpenes and isoprene in laboratoryexperiments. Our results show that the ozonolysis of endocy-clic monoterpenes, such as α-pinene and limonene, producesELVOC very rapidly and with much greater yields than the OHradical-initiated oxidation. Ozonolysis of exocyclic or acyclicmonoterpenes, such as β-pinene, myrcene, and isoprene, werealso found to produce ELVOC, but with significantly lowermolar yields. Remarkably, we also found a direct, although weak,source of ELVOC from the OH oxidation of acyclic and exo-cyclic terpenes and isoprene, not reported previously. Oxidationof isoprene leads to only marginal ELVOC formation.We implemented our laboratory findings into a global mod-

eling framework and showed that biogenic SOA formation ingeneral, and ELVOC in particular, plays a crucial role in at-mospheric CCN production. ELVOC, originating mainly frommonoterpene oxidation, enhance atmospheric new particle for-mation and growth in most continental regions, thereby in-creasing CCN concentrations, especially at high values of cloudsupersaturation. Isoprene-derived SOA tends to suppresses newparticle formation and early growth, at the same time assistingthe growth of sub-CCN-size primary particles to CCN. Althoughdifferent monoterpenes appear to produce ELVOC at very dif-ferent efficiencies, the overall effect of speciated monoterpeneemissions on the global CCN budget appears to be moderate (afew percent in maximum). This latter effect may, however, beunderestimated due to the limited variability in monoterpenespeciation applied in the current biogenic emission model. Amore accurate assessment requires more-realistic representa-tions of monoterpene speciation variability based on systematicobservations in key ecosystems and an approach to extend theplant type scheme used in the biogenic VOC emission model.

Our results are of direct importance for understanding andmodeling the connections among BVOC emissions, atmosphericoxidation, new particle formation, and secondary CCN pro-duction. High yields of ELVOC from atmospheric oxidation ofBVOC enhance the formation of new aerosol particles and theirsubsequent growth to CCN. However, various compounds thatare more volatile than ELVOC and that originate from BVOCoxidation may enhance atmospheric CCN production, providedthat sub-CCN-size primary carbonaceous particles are availableto uptake these compounds. Our study demonstrates the needfor a better quantification of BVOC emission patterns, as wellas improvement of modeling approaches to properly accountfor the capabilities of BVOC to produce secondary products ofdifferent volatility.

Materials and MethodsExperiments were performed at atmospheric pressure in the TROPOS-LFTlaminar flow glass tube (i.d. 8 cm, length 505 cm) (37) at constant temper-ature (293 ± 0.5 K) and relative humidity (25%) and 40-s reaction time.Purified synthetic air, premixed with selected terpene, was introduced at thetop of the flow tube with a flow rate of 30 L min−1 (standard conditions fortemperature and pressure, STP). Ozone was diluted in the carrier gas, and itwas mixed with the main stream using nozzles. OH radicals were producedfrom terpene ozonolysis. Terpene concentrations were monitored withproton transfer reaction mass spectrometry (PTR-MS) (High-Sensitivity PTR-MS; Ionicon) (38). ELVOC concentrations were measured using a nitrate ion-based chemical ionization atmospheric pressure interface time-of-flight (CI-APi-TOF) mass spectrometer (39–41) with a sample flow rate of 10 L·min−1

(STP). The CI-APi-TOF calibration was based on sulfuric acid detection, andthe resulting calibration coefficient was used to calculate ELVOC concen-trations. Measurement uncertainty is estimated to be −50%/+100%, arisingfrom the calibration and from the possibility that the transmission of theELVOC NO3

− may differ from the transmission of the HSO4− used for cali-

bration. Also, the charging probability may be different for the highly oxi-dized compounds than for H2SO4 (which is ionized at the collision limit); forfurther details, see ref. 21 and SI Appendix.

ECHAM5-HAM (42) is an aerosol-climate model originally developed atMax Planck Institute, Hamburg. We used the version ECHAM5.5-HAM2 (43).The Hamburg Aerosol Module (HAM) describes aerosol population with foursoluble and three insoluble log-normal modes, including sulfate, organiccarbon, black carbon, dust, and sea salt. In the original ECHAM5-HAMmodel, the simulated gas-phase compounds are SO2, DMS, and sulfate. Inthis study, we extended the SOA formation model developed in Makkonenet al. (44) to include both kinetic condensation to Fuchs-corrected surfacearea (condensation sink) and partitioning according to preexisting organicmass. Three gas-phase tracers were included for BVOC: isoprene, endocyclic,and other monoterpenes. Included endocyclic monoterpenes are α-pinene,limonene, α-phellandrene, β-phellandrene, 3-carene, terpinolene, α-terpi-nene, and γ-terpinene. Compounds in the “other monoterpenes” group arecamphene, β-pinene, myrcene, and t-β-ocimene. The reaction rates of SOAprecursors with O3, OH, and NO3 are shown in SI Appendix, Table S3. Oxi-dant concentrations are prescribed monthly averages as in Stier et al. (42);hence the model does not include chemistry feedbacks via BVOC emissions.Simulations include organic vapors in the nucleation process according toequation 18 in Paasonen et al. (45). The growth from nucleation to 3 nm iscalculated according to Kerminen and Kulmala (46) assuming growth byELVOC and sulfuric acid. All simulations are run for 1 y after 3-mo spinupusing nudging toward year 2000 European Center for Medium-RangeWeatherForecast-40 (ERA-40) reanalysis fields.

The emissions of BVOC were monthly averages calculated offline bythe Model of Emissions of Gases and Aerosols from Nature version 2.1(MEGAN2.1) (24), which provides a simplified mechanistic approach for es-timating global distributions of biogenic hydrocarbon emissions as a func-tion of land cover and multiple environmental variables. The input variablesdescribed by Sindelarova et al. (47) were used to drive the simulation usedfor this study. The simulation used global average emission factors for plantfunctional types, which greatly limit the variability of the relative contri-bution of specific terpenes, and the actual variability is likely to bemuch greater.

ACKNOWLEDGMENTS. We thank K. Pielok, R. Gräfe, and A. Rohmer fortechnical assistance, Matti Rissanen for valuable discussions, Pontus Roldinfor SIMPOL and Nannoolal method volatility calculations, and the tofToolsteam for proving a toolbox for data analysis. This work was partly funded by

Jokinen et al. PNAS | June 9, 2015 | vol. 112 | no. 23 | 7127

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Dec

embe

r 28

, 202

0

the Pan-European Gas-Aerosol-Climate Interaction Study (PEGASOS) andImpact of Biogenic versus Anthropogenic emissions on Clouds and Cli-mate: towards a Holistic Understanding (BACCHUS) projects [funded bythe European Commission under the Framework Program 7 (FP7-ENV-

2010-265148 and FP7-ENV-2013-603445)], the Academy of Finland (Centerof Excellence, Grants 1118615 and 251427), and the European ResearchCouncil [Atmospheric nucleation: from molecular to global scale (ATMNUCLE)](Grant 227463).

1. Merikanto J, Spracklen DV, Mann GW, Pickering SJ, Carslaw KS (2009) Impact of nu-cleation on global CCN. Atmos Chem Phys 9(21):8601–8616.

2. Kerminen V-M, et al. (2012) Cloud condensation nuclei production associated withatmospheric nucleation: A synthesis based on existing literature and new results.Atmos Chem Phys 12(24):12037–12059.

3. Kazil J, et al. (2010) Aerosol nucleation and its role for clouds and Earth’s radiativeforcing in the aerosol-climate model ECHAM5-HAM. Atmos Chem Phys 10(22):10733–10752.

4. Makkonen R, et al. (2012) Air pollution control and decreasing new particle formationlead to strong climate warming. Atmos Chem Phys 12(3):1515–1524.

5. Ghan SJ (2013) A simple model of global aerosol indirect effects. J Geophys Res118(12):6688–6707.

6. Carslaw KS, et al. (2013) Large contribution of natural aerosols to uncertainty in in-direct forcing. Nature 503(7474):67–71.

7. Boucher O, et al. (2013) Clouds and aerosols. Climate Change 2013: The Physical Sci-ence Basis. Contribution of Working Group I to the Fifth Assessment Report of theIntergovernmental Panel on Climate Change, eds Stocker TF, et al. (Cambridge UnivPress, Cambridge, UK).

8. Tunved P, et al. (2006) High natural aerosol loading over boreal forests. Science312(5771):261–263.

9. Jimenez JL, et al. (2009) Evolution of organic aerosols in the atmosphere. Science326(5959):1525–1529.

10. Kulmala M, et al. (2013) Direct observations of atmospheric aerosol nucleation. Sci-ence 339(6122):943–946.

11. Paasonen P, et al. (2013) Warming-induced increase in aerosol number concentrationlikely to moderate climate change. Nat Geosci 6(6):438–442.

12. Riipinen I, et al. (2011) Organic condensation: A vital link connecting aerosol for-mation to cloud condensation nuclei (CCN) concentrations. Atmos Chem Phys 11(8):3865–3878.

13. Makkonen R, et al. (2012) BVOC-aerosol-climate interactions in the global aerosol-climate model ECHAM5.5-HAM2. Atmos Chem Phys 12(21):10077–10096.

14. D’Andreae SD, et al. (2013) Understanding global secondary organic aerosol amountand size-resolved condensational behavior. Atmos Chem Phys 13(7):11519–11534.

15. Riipinen I, et al. (2012) The contribution of organics to atmospheric nanoparticlegrowth. Nat Geosci 5(7):453–458.

16. Kulmala M, Toivonen A, Mäkelä J (1998) Analysis of the growth of nucleation modeparticles observed in boreal forest. Tellus Ser B 50(5):449–462.

17. Donahue NM, Epstein SA, Pandis SN, Robinson AL (2011) A two-dimensional volatilitybasis set – Part 1: Organic aerosol mixing thermodynamics. Atmos Chem Phys 11(7):3303–3318.

18. Ehn M, et al. (2012) Gas phase formation of extremely oxidized pinene reactionproducts in chamber and ambient air. Atmos Chem Phys 12(11):5113–5127.

19. Hallquist M (2009) The formation, properties and impact of secondary organicaerosol: Current and emerging issues. Atmos Chem Phys 9(14):5155–5236.

20. Ehn M, et al. (2014) A large source of low-volatility secondary organic aerosol. Nature506(7489):476–479.

21. Jokinen T, et al. (2014) Rapid autoxidation forms highly oxidized RO2 radicals in theatmosphere. Angew Chem Int Ed Engl 53(52):14596–14600.

22. Rissanen MP, et al. (2014) The formation of highly oxidized multifunctional productsin the ozonolysis of cyclohexene. J Am Chem Soc 136(44):15596–15606.

23. Crounce JD (2013) Autoxidation of organic compounds in the atmosphere. J PhysChem Lett 4(20):3513–3520.

24. Guenther AB, et al. (2012) The Model of Emissions of Gases and Aerosols from Natureversion 2.1 (MEGAN2.1): An extended and updated framework for modeling biogenicemissions. Geosci Model Dev 5(6):1471–1492.

25. Tarvainen V, Hakola H, Rinne J, Hellén H, Haapanala S (2007) Towards a compre-hensive emission inventory of terpenoids from boreal ecosystems. Tellus Ser B 59(3):526–534.

26. Bäck J, et al. (2012) Chemodiversity of a Scots pine stand and implications for terpeneair concentrations. Biogeosciences 9(2):689–702.

27. Kuhn U, et al. (2007) Isoprene and monoterpene fluxes from Central Amazonianrainforest inferred from tower-based and airborne measurements, and implicationson the atmospheric chemistry and the local carbon budget. Atmos Chem Phys 7(11):2855–2879.

28. Kuhn U, et al. (2004) Seasonal differences in isoprene and light-dependent mono-terpene emission by Amazonian tree species. Global Change Biol 10(5):663–682.

29. Constable JVH, Litvak ME, Greenberg JP, Monson RK (1999) Monoterpene emissionfrom coniferous treens in response to elevated CO2 and climate warming. GlobalChange Biol 5:255–267.

30. Peñuelas J, Staudt M (2010) BVOCs and global change. Trends Plant Sci 15(3):133–144.31. Arneth A, et al. (2011) Global terrestrial isoprene emission models: Sensitivity to

variability in climate and vegetation. Atmos Chem Phys 11(15):8037–8052.32. Pankow JF, Ashew WE (2008) SIMPOL.1: A simple group contribution method for

predicting vapor pressures and enthalpies of vaporization of multifunctional organiccompounds. Atmos Chem Phys 8(10):2773–2796.

33. Nannoolal Y, Rareya J, Ramjugernatha D, Cordesb W (2008) Estimation of purecomponent properties. Part 1. Estimation of the normal boiling point of non-elec-trolyte organic compounds via group contributions and group interactions. FluidPhase Equilib 226:45–63.

34. Tsikaridis K, et al. (2014) The AeroCom evaluation and intercomparison of organicaerosol in global models. Atmos Chem Phys 14(19):10845–10895.

35. Spracklen DV, et al. (2011) Aerosol mass spectrometer constraint on the global sec-ondary organic aerosol budget. Atmos Chem Phys 11(23):12109–12136.

36. Mann GW, et al. (2014) Intercomparison and evaluation of global aerosol micro-physical properties among AeroCom models of a range of complexity. Atmos ChemPhys 14(9):4679–4713.

37. Berndt T, Böge O, Stratmann F, Heintzenberg J, Kulmala M (2005) Rapid formation ofsulfuric acid particles at near-atmospheric conditions. Science 307(5710):698–700.

38. Lindinger W, Jordan A (1998) Proton-transfer-reaction mass spectrometry (PTR–MS):On-line monitoring of volatile organic compounds at pptv levels. Chem Soc Rev 27(5):347–375.

39. Eisele F, Tanner D (1993) Measurement of the gas phase concentration of H2SO4 andmethane sulfonic acid and estimates of H2SO4 production and loss in the atmosphere.J Geophys Res 98(D5):9001–9010.

40. Junninen H, et al. (2010) A high-resolution mass spectrometer to measure atmo-spheric ion composition. Atmos Meas Tech 3(4):1039–1053.

41. Jokinen T, et al. (2012) Atmospheric sulfuric acid and neutral cluster measurementsusing CI-APi-TOF. Atmos Chem Phys 12(9):4117–4125.

42. Stier P, et al. (2005) The aerosol-climate model ECHAM5-HAM. Atmos Chem Phys 5(4):1125–1156.

43. Zhang K, et al. (2012) The global aerosol-climate model ECHAM-HAM, version 2:Sensitivity to improvements in process representations. Atmos Chem Phys 12(19):8911–8949.

44. Makkonen R, et al. (2009) Sensitivity of aerosol concentrations and cloud propertiesto nucleation and secondary organic distribution in ECHAM5-HAM global circulationmodel. Atmos Chem Phys 9(5):1747–1766.

45. Paasonen P, et al. (2010) On the roles of sulphuric acid and low-volatility organicvapours in the initial steps of atmospheric new particle formation. Atmos Chem Phys10(22):11223–11242.

46. Kerminen V-M, Kulmala M (2002) Analytical formulae connecting the “real” and the“apparent” nucleation rate and the nuclei number concentration for atmosphericnucleation events. J Aerosol Sci 33(4):609–622.

47. Sindelarova K, et al. (2014) Global data set of biogenic VOC emissions calculated bythe MEGAN model over the last 30 years. Atmos Chem Phys 14(17):9317–9341.

7128 | www.pnas.org/cgi/doi/10.1073/pnas.1423977112 Jokinen et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 28

, 202

0